At present, 3D surveillance videos have been widely used in the field of video surveillance. A 3D surveillance video can display the events happened in a monitored area in a stereoscopic and panoramic manner, and realize the organic combination of the overall view and the detail control.
The method for generating a 3D surveillance video generally includes: establishing a 3D scene model of a monitored area; fusing videos of this monitored area from various fields of view within the same period to the 3D scene model so as to generate a 3D panoramic surveillance video for this monitored area.
In practical applications, it is often the case that a plane in the 3D surveillance video generated from the fusion of a video and a 3D model is uneven, which results in poor fusion result. For example, when establishing a 3D model, a certain ground is a lawn. However, this ground has become a flat floor when taking a video. In this case, the resulting ground will be very uneven from the fusion of this 3D model and this video, resulting in a very poor visual impression.
For such a case, the typical solution is to modify the 3D model so that the ground generated after the fusion is a flat ground. However, this requires a frequent modification to the 3D model very often when fusing the 3D model with different videos, while the process of modifying a 3D scene model is very complex, resulting in both high labor cost and high time cost.